Software for Structural Bioinformatics of Nucleic Acid

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Methods for macromolecular structure determination cannot keep pace with the exponential growth of biological sequence databases. The goal of this proposal is to develop a new software platform for structural bioinformatics of nucleic acids that will allow analysis of existing structures and accurate prediction of 3D structures. This will be achieved in four specific aims: Aim 1: Write software for homology modeling of nucleic acids. Software will be written to "thread" a sequence whose 3D structure is unknown into a known template structure. This methodology will leverage genome sequencing projects by allowing the automated 3D structure prediction of many functional RNAs. Aim 2: Write software for de novo 3D structure prediction of nucleic acids. Secondary structures from a dynamic programming algorithm will be converted into rough 3D structures using a novel BUILDER algorithm that utilizes a motif database. Classical molecular dynamics simulations will be used to refine the structures. The result will be accurate de novo 3D structure predictions of RNA and DNA. Aim 3: Develop an extended forcefield and optimization algorithms for nucleic acids. The AMBER forcefield will be extended to include parameters for the >100 modified nucleotides that occur in natural RNAs. The forcefields will also be extended to include pseudo-potential terms for gap penalties, solution trends in DNA and RNA folding thermodynamics, and experimental restraints. This modified forcefield will be used to rank predicted 3D structures. New geometry optimization algorithms will also be developed. Aim 4: A systematic validation of the quality of predictions of the software produced by Aims 1,2 and 3 will be performed. The software will be used to predict all the published 3D structures determined by X-ray crystallography or NMR. A blind test will be performed by inviting structural biologists to submit unpublished structures for prediction by our algorithms (similar to CASP for proteins).